A-SIS Deduplication on Netapp.

[joachims1]

Is deduplication suitable for use on a production vmfs volume? Does anybody know how dedup impacts on performance of a filer? We have a FAS3020 and priceing for dedup option is not to bad.

Have anybody tried this?

[bastop]

I heard ASIS pegs filers, even FAS6000 series. I heard this from several sources, including competitors (ie Data Domain).

I can't understand why you would want to run ASIS. I thought it was more meant for a netbackup type volume or NAS deduplication.

I can't see it being useful for block level access like your vmware luns.

[joachims1]

If performance is bad i certanly wont use it. I only heard that you could save big on diskspace. Up to 80% is what i heard.

[bastop]

Is Netapp saying specifically LUNs / block level stuff you can save? I would certainly try it out with a temp license key first if you are looking at it. Netapp shelves aren't cheap, but temp license keys are easily available from your rep.

[siglert]

For the FAS systems the DATA On-TAP software is much better. I just did an engagement for a client for disaster Recovery an the Data replication works like a charm.

[dedupe]

NetApp doesn't recommend running A-SIS on production vmfs volumes. They are referring to backup copies of clone images, thats where they get the big space savings with deduplication. BTW A-SIS doesn't "peg" the filer, that sounds like FUD from competitors.

[dalepa]

We run A-SIS on 4 TB vmware volumes on our snapmirror destination and we are getting 50-70% deduplication rates.

On our asis3 volume (see below), we are cramming 7.2TB into a 4TB volume and we still have 1.4TB free. Asis3 also contains 800G in snapshots. asis3 holds 613 vm guests with a mix of XP, Win2.3K and Redhat servers.

Observations:

A-SIS is configurable on a per volume basis.

A-SIS is currently limited to 4TB

If you keep like OS's on a volume you will get higher deduplication rates.

The initial deduplication of a large 4TB volume will take all weekend, and does put a strain on the CPU, however this is a one-time event. After the initial deduplication, only delta's are deduped and the process only runs an hour or so each night. You can schedule the dedups to run on weekends if you like...

A-SIS is supported on your primary filer as of version 7.2.2. Since A-SIS is only about a year old and we have plenty of space, we decided to wait before running on the primary filer. We have tested running on the primary filer with live VMs without any issues.

A-SIS should be supported at the Aggregate level later this year, and the 16TB Aggregate limit will be raised!

We also deduped an oracle volume and got a 30% deduplication rate.

Deduplication is the NEXT big thing in storage, and Netapp is leading the pack...

r200> df -sg

Filesystem used saved %saved

/vol/asis1/ 1864GB 1639GB 47%

/vol/asis2/ 172GB 340GB 66%

/vol/asis3/ 2653GB 4594GB 63%

[vstewart]

There's a few points of misinformation on this topic

NetApp A-SIS or Data Deduplication has volume size limits based on the platform which is hosting the data.

For example a FAS6070 has a maximum individual A-SIS enabled volume size of 14TB

Volumes are initialized with A-SIS, which will de-dupe the content, and subsequent de-dupe updates are scheduled. The performance overhead for enabling A-SIS is up to a 10% write performance penalty; however, customers have seen read performance improve. Note your mileage may vary based on solution architecture.

A-SIS today does have some meta data overhead which is stored in the volume, so it may be advantageous to run the incremental updates on percentage of data change schedule rather than nightly.

A-SIS provides the same storage savings on FCP, iSCSI, NFS, VMFS, RDMs, & VMDKs. The difference in storage options is with NFS the savings are seen by the VMware admins, with LUNs the savings are only seen by the storage admins.

Data De Duplication makes allot of sense with VMware as most Vms are deployed from (or copied from) templates. Your sharing server hardware why not share duplicate data?